[lucene] 09/18: SOLR-12923: fix SimClusterStateProvider to use lock.lockInterruptibly() exclusively, and make SimCloudManager's Callable checks tollerant of Callables that may have failed related to interrupts w/o explicitly throwing InterruptedException

[dw...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

dweiss pushed a commit to branch branch_8_0
in repository https://gitbox.apache.org/repos/asf/lucene.git

commit 4552eeddc613a5e143c97963b504102c6f41af46
Author: Chris Hostetter <ho...@apache.org>
AuthorDate: Fri Mar 15 12:06:53 2019 -0700

    SOLR-12923: fix SimClusterStateProvider to use lock.lockInterruptibly() exclusively, and make SimCloudManager's Callable checks tollerant of Callables that may have failed related to interrupts w/o explicitly throwing InterruptedException
    
    (cherry picked from commit 1a54c6b19db9dcb1081e43614bf479e0ac7bf177)
---
 .../cloud/autoscaling/sim/SimCloudManager.java     | 15 ++++++++++++++
 .../autoscaling/sim/SimClusterStateProvider.java   | 23 +++++++++++++---------
 2 files changed, 29 insertions(+), 9 deletions(-)

diff --git a/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimCloudManager.java b/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimCloudManager.java
index fda1892..7ce4534 100644
--- a/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimCloudManager.java
+++ b/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimCloudManager.java
@@ -967,6 +967,21 @@ public class SimCloudManager implements SolrCloudManager {
         log.warn("Callable interupted", ignored);
         throw ignored;
       } catch (Throwable t) {
+        // be forgiving of errors that occured as a result of interuption, even if
+        // the inner Callable didn't realize it...
+        if (Thread.currentThread().isInterrupted()) {
+          log.warn("Callable interupted w/o noticing", t);
+          throw t;
+        }
+        Throwable cause = t;
+        while ((cause = cause.getCause()) != null) {
+          if (cause instanceof InterruptedException) {
+            log.warn("Callable threw wrapped InterruptedException", t);
+            throw t;
+          }
+        }
+
+        // in all other situations, this is a problem that should be tracked in the failCounter
         failCounter.incrementAndGet();
         log.error("Callable failed", t);
         throw t;
diff --git a/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimClusterStateProvider.java b/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimClusterStateProvider.java
index 7e738b9..3d6bece 100644
--- a/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimClusterStateProvider.java
+++ b/solr/core/src/test/org/apache/solr/cloud/autoscaling/sim/SimClusterStateProvider.java
@@ -18,6 +18,7 @@
 package org.apache.solr.cloud.autoscaling.sim;
 
 import java.io.IOException;
+import java.lang.InterruptedException;
 import java.lang.invoke.MethodHandles;
 import java.util.ArrayList;
 import java.util.Arrays;
@@ -190,7 +191,7 @@ public class SimClusterStateProvider implements ClusterStateProvider {
    * @param initialState initial cluster state
    */
   public void simSetClusterState(ClusterState initialState) throws Exception {
-    lock.lock();
+    lock.lockInterruptibly();
     try {
       collProperties.clear();
       sliceProperties.clear();
@@ -2148,20 +2149,24 @@ public class SimClusterStateProvider implements ClusterStateProvider {
   @Override
   public ClusterState getClusterState() throws IOException {
     ensureNotClosed();
-    lock.lock();
     try {
-      Map<String, DocCollection> states = getCollectionStates();
-      ClusterState state = new ClusterState(clusterStateVersion, liveNodes.get(), states);
-      return state;
-    } finally {
-      lock.unlock();
+      lock.lockInterruptibly();
+      try {
+        Map<String, DocCollection> states = getCollectionStates();
+        ClusterState state = new ClusterState(clusterStateVersion, liveNodes.get(), states);
+        return state;
+      } finally {
+        lock.unlock();
+      }
+    } catch (InterruptedException e) {
+      throw new IOException(e);
     }
   }
 
   // this method uses a simple cache in collectionsStatesRef. Operations that modify
   // cluster state should always reset this cache so that the changes become visible
-  private Map<String, DocCollection> getCollectionStates() throws IOException {
-    lock.lock();
+  private Map<String, DocCollection> getCollectionStates() throws IOException, InterruptedException {
+    lock.lockInterruptibly();
     try {
       Map<String, DocCollection> collectionStates = collectionsStatesRef.get();
       if (collectionStates != null) {